Objectives. We know little about the variation in diagnosis and management of bronchiolitis. The objectives of this study were (1) to document variations in treatment and diagnostic approaches, lengths of stay (LOSs), and readmission rates and (2) to determine which potentially modifiable process of care measures are associated with longer LOSs and antibiotic usage.
Methods. We used the Pediatric Health Information System, which includes demographic, diagnostic, and detailed patient-level data on 30 large children's hospitals. We examined infants who were younger than 1 year and hospitalized for bronchiolitis (October 2001–September 2003). Multivariate analysis of variance was used to determine whether the variance in the outcomes was hospital related after controlling for other covariates. Linear regression was used to model predictors of increased LOS. Logistic regression was used to model antibiotic usage. Analyses were stratified by age group (<3 months and 3–11 months).
Results. A total of 17397 patients were included in the analysis. The mean LOS was 2.97 days; 72% of patients received chest radiographs, 45% received antibiotics, and 25% received systemic steroids. The mean LOS varied considerably across hospitals (range: 2.40–3.90 days), and hospital remained a significant contributor to LOS variation after controlling for our covariates. Variations in the use of diagnostic tests and medications as well as readmission rates also existed and also remained significant after controlling for covariates. The factors associated with the greatest increases in LOS in the regression analyses included higher severity scores and use of antibiotics, bronchodilators, and corticosteroids. The strongest predictors of antibiotic use in the logistic regression analyses were higher severity scores and receipt of a blood or cerebrospinal fluid culture. Receiving a chest radiograph was a significant predictor of antibiotic use in older but not younger infants.
Conclusions. Considerable, unexplained variation exists in the inpatient management of bronchiolitis. The development of national guidelines and controlled trials of new therapies and different management approaches are indicated.
Most commonly caused by respiratory syncytial virus, bronchiolitis is a clinical syndrome characterized by wheezing, hypoxia, and tachypnea and is the most common lower respiratory tract infection that afflicts infants.1,2 Affecting almost all children within the first 2 years of life, bronchiolitis remains 1 of the most frequent causes of hospitalization in this age group, and the rates of hospitalization for bronchiolitis have been increasing.1,3 Despite how common bronchiolitis is, we have remarkably few effective therapies.4–7
Routine use of radiologic studies does not seem warranted as the findings are nonspecific and frequently lead to the use of antibiotics without demonstrable benefit.8–10 β2-Agonist bronchodilators have also proved to be largely ineffective.5,11 Racemic epinephrine, in contrast, has shown promising if transient benefits in a small subset of patients.11–16 Systemic steroids may be associated with small decreases in lengths of stay (LOSs) and symptoms and may be especially beneficial in more severe cases, although these results remain controversial.4,11,17–19 Finally, although there is some evidence that inhaled steroids are effective in the treatment of episodic wheezing that can follow acute bronchiolitis,20 there is little evidence that they are an effective treatment for the acute condition.21,22
Evidence-based clinical pathways developed and implemented at some institutions have been associated with the decreased use of rapid virological diagnostic tests, antibiotics, and chest radiographs, as well as shorter LOSs.23–25 The success of other institutions in implementing a more evidence-based approach to bronchiolitis management to decrease the unnecessary use of diagnostic tests and medication is unknown. We conducted a large retrospective descriptive study of infants who were admitted with bronchiolitis to large children's hospitals across the United States. Our goals were twofold: (1) to document variations in treatment, diagnostic approaches, LOSs, and readmission rates and (2) to determine which potentially modifiable process measures of care were associated with longer LOSs and antibiotic usage.
We used the Pediatric Health Information System database developed by the Child Health Corporation of America, which includes demographic and diagnostic data on 36 freestanding, noncompeting children's hospitals. For 30 of these, a more comprehensive data set that includes medications and diagnostic testing is available.26,27 This analysis is limited to patients for whom the expanded data were available. The database uses the International Classification and Clinical Services codes28 to map hospital-specific charge codes at the patient level to categorical variables across all hospitals. The database also includes diagnoses in International Classification of Diseases, Ninth Revision format, as well as All-Patient Refined Diagnosis Related Groups (APR-DRGs), version 15. The study protocol was reviewed and approved by the Children's Hospital and Regional Medical Center Institutional Review Board.
We included infants who were younger than 1 year and hospitalized for bronchiolitis with discharge dates between October 1, 2001, and September 30, 2003. We restricted the analysis to infants who had both a primary discharge diagnosis of bronchiolitis (International Classification of Diseases, Ninth Revision code of either 466.11 or 466.19) and an APR-DRG of bronchiolitis/asthma (141).29 When an infant was hospitalized more than once for bronchiolitis during the study period, the first hospitalization was included in the primary analysis.
In a descriptive analysis, we examined variability across hospitals in LOS, diagnostic testing (including chest radiographs, rapid virological test for respiratory syncytial virus and influenza, bacterial cultures [from blood, cerebrospinal fluid, nasopharynx, and urine]), medications used [parenteral or oral antibiotics, racemic epinephrine, inhaled steroids, and systemic steroids], and readmission rates. We did not examine bronchodilator use because despite the limited data on their efficacy, they are frequently used on a trial basis because a small proportion of children are deemed responsive. Accordingly, bronchodilator use, measured dichotomously (used or not used), yields little information.
We defined readmission as a repeat admission for bronchiolitis that occurred within 3 days of the index discharge date. There is no widely accepted definition of what constitutes a readmission for the same illness. We chose 3 days somewhat arbitrarily but believing that it has clinical face validity for representing the same episode of illness and may reflect premature discharge. With the exception of LOS, which was measured continuously, all of our other outcome variables were measured dichotomously.
The following additional variables were included as covariates in the regressions: age (in months), gender, Medicaid status, severity classification (derived from the APR-DRG Severity of Illness Guidelines),30 and month of admission (to adjust for seasonal trends).
χ2 tests were used to compare categorical variables, and t tests were used to compare continuous ones. Multivariate analysis of variance was used to determine whether hospital was a significant contributor to the variance in the outcomes after controlling for other covariates.
We also conducted 2 regression analyses: a linear regression examining LOS and a logistic regression examining the usage of any antibiotic (parenteral or oral). Although the distribution of LOS is highly skewed, our large sample size made linear regression appropriate.31 The β coefficients in linear regression can be interpreted as the additional days in average LOS associated with a 1-unit change in each predictor.
The regression analyses were clustered on hospital, to account for the decreased variability within hospitals as compared with between hospitals.32,33 As treatment of very young infants with bronchiolitis can, for clinical reasons, vary from that for older infants, we stratified the regressions by age (<3 months vs 3–11 months). To protect the integrity of the participating hospitals, all results are presented with hospitals de-identified. All analyses were conducted using Stata 8.0 (Stata Corp, College Station, TX).
A total of 17397 patients were included in the analysis. The study sample is described in detail in Table 1. Briefly, the mean age was 3.96 months, and 59% were male. The mean LOS was 2.97 days; 72% received chest radiographs, 45% received antibiotics, and 25% received systemic steroids.
The mean LOS varied considerably across hospitals (range: 2.40–3.90 days; Fig 1), and hospital remained a significant contributor to LOS variation after controlling for our covariates (P < .001). There was significant variation across hospitals in use of diagnostic tests (Fig 2) and medication (Fig 3), and, again, these differences remained significant after controlling for covariates (P < .001). For example, the use of any antibiotics varied from 28% to 62% across hospitals, and the use of chest radiographs varied from 38% to 89%. There were also significant differences across hospitals in readmission rates (Fig 1), varying from 0% to 2.7% across hospitals, and they remained significant after controlling for covariates, including LOS (P < .001). In addition, longer LOSs were associated with decreased likelihood of readmission. Compared with children who were discharged with an LOS of 1 day, children with LOS of 2 days (odds ratio [OR]: 0.49; 95% confidence interval [CI]: 0.37 to 0.66), 3 days (OR: 0.41; 95% CI: 0.25 to 0.68), or 4 or more days (OR: 0.29; 95% CI: 0.19 to 0.46), all were significantly less likely to be readmitted.
In the linear regression analysis for children ≥3 months, the following were associated with increased LOSs: having obtained a chest radiograph, blood culture, or urine culture, and receiving parenteral or oral antibiotics or systemic steroids. Age (in months) was associated with decreased LOS. The results for children who were younger than 3 months are largely similar to those who were≥3 months of age (Table 2).
In the logistic regression analysis in children who were ≥3 months, the following were associated with increased likelihood of antibiotic use: having obtained a chest radiograph, a blood culture, cerebrospinal fluid culture, or urine culture. For younger infants, the same analysis yielded similar results. Use of viral testing was associated with decreased likelihood of antibiotics, although it did not reach statistical significance (OR: 0.91; 95% CI: 0.81 to 1.01). In a separate analysis of all children who were younger than 1 year, the point estimate remained the same and the result was statistically significant (data not shown). Complete results are presented in Table 3.
We found significant and wide variation in LOS, readmission rates, treatment approaches, and the use of diagnostic tests for inpatient bronchiolitis. The variation between hospitals remained significant even after controlling for multiple potential confounding factors. To our knowledge, this is the largest study to compare the treatment of infants who are hospitalized for bronchiolitis, and it demonstrates that there is ample room for improvement nationally in the quality of care provided. Decreasing LOS and unnecessary medication and test utilization is also supportive of pediatric patient safety initiatives.
Some of our findings warrant specific comment. Chest radiographs were associated with an increased likelihood of using antibiotics. The findings of bronchiolitis are nonspecific, and the presence of atelectasis, which may not be distinguishable from consolidation, may lead clinicians to treat presumptively even if unnecessarily. Reducing the use of radiographs in children with bronchiolitis does seem warranted and has been shown by others to be associated with decreased probability of antibiotic use.23,24 Moreover, if virologic testing leads to decreased use of antibiotics, then it might very well be a cost-effective approach. Although a formal cost-effectiveness analysis is beyond the scope of this article, the marginal increased cost associated with such testing likely could be more than offset with the costs of administering antibiotics.
Although the reasons for the variations that we observed in diagnostic and therapeutic approaches are unknown, given that we have endeavored to adjust for a number of potentially explanatory variables including age, comorbidities, etc, one may wonder what accounts for them. One possible explanation is institutional cultural differences (viz, some places “get radiographs,” whereas others “do not”). Radiologic and virologic tests, although costly, are relatively benign. The variation in treatments, however, is more problematic. Data from well-conducted randomized trials have shown that inhaled steroids do not work acutely.21,22 That a proven ineffective therapy is used in as many as 14% of bronchiolitis patients in some hospitals begs additional study, as does the frequent and wide variation in the use of antibiotics.
There are some limitations to this study that warrant mention. First, our data were collected retrospectively and are cross-sectional. Accordingly, it is not possible to determine whether significant selection bias was present in the use of diagnostic testing, and causal inferences are not appropriate. That is, more severely affected patients may have been more likely to have chest radiographs performed and in fact more likely to have a bacterial superinfection. For example, Wang et al34 found considerable variability in the percentage of children who were admitted with bronchiolitis and had hypoxemia. However, that study included all children who were admitted to the hospital, including those for whom bronchiolitis would not have been the principal diagnosis. It seems highly improbable, however, that variations in the severity of patients alone can account for the wide variations in the use of diagnostic tests, including chest radiographs, especially because our analysis controlled for APR-DRG–derived severity scores, which take into account secondary diagnoses that would be associated with a true need for antibiotic treatment (eg, pneumonia). Second, although we attempted to control for multiple sources of potential confounding in our analyses, it is possible that some residual confounding remains. Whether this systematically biased our results would depend on whether that variation was associated with particular institutions, something that we have no a priori reason to suspect.
Despite these limitations, there are some important implications to our findings. First, there is a need to study the inpatient management of common pediatric conditions. Such widespread variations in the process and outcome measures of care, particularly in situations in which a sound evidence base exists, argues for national quality improvement efforts directed at pediatric inpatient care. Second, better therapeutic approaches are needed for bronchiolitis care. Whether these are at the system level in the form of guidelines or pathways or are in the form of new treatment modalities, robust prospective studies are in order. Ideally, such trials would be accomplished with the collaboration of several institutions, which would make the trials more feasible and the findings more generalizable. For example, it was estimated that a trial of systemic steroids in the treatment of broncholiths would require 728 patients to be powered adequately,4 yet smaller, underpowered trials continue to be performed.35–38 The definitive, large, randomized, controlled trial exceeds the ability of a single institution to conduct but is well within the means of several collaborating hospitals. In the interim, on the basis of the results presented here, individual institutions should direct resources at assessing and potentially improving their inpatient bronchiolitis care.
We are grateful to the Children's Hospital and Regional Medical Center of Seattle for ongoing support of quality improvement research.
- Accepted August 4, 2004.
- Address correspondence to Dimitri Christakis, MD, MPH, Child Health Institute, 6200 NE 74th St, Suite 210, Seattle, WA 98115-8160. E-mail:
No conflict of interest declared.
- ↵Parrott RH, Kim HW, Arrobio JO, et al. Epidemiology of respiratory syncytial virus infection in Washington, D.C. II. Infection and disease with respect to age, immunologic status, race and sex. Am J Epidemiol.1973;98 :289– 300
- ↵Langley JM, LeBlanc JC, Smith B, Wang EE. Increasing incidence of hospitalization for bronchiolitis among Canadian children, 1980–2000. J Infect Dis.2003;188 :1764– 1767
- ↵Garrison MM, Christakis DA, Harvey E, Cummings P, Davis RL. Systemic corticosteroids in infant bronchiolitis: a meta-analysis. Pediatrics.2000;105(4) . Available at: www.pediatrics.org/cgi/content/full/105/4/e44
- ↵Flores G, Horwitz RI. Efficacy of β2-agonists in bronchiolitis: a reappraisal and meta-analysis. Pediatrics.1997;100(suppl) :233– 239
- Mallory MD, Shay DK, Garrett J, Bordley WC. Bronchiolitis management preferences and the influence of pulse oximetry and respiratory rate on the decision to admit. Pediatrics.2003;111(1) . Available at: www.pediatrics.org/cgi/content/full/111/1/e45
- ↵Hall CB. Managing bronchiolitis and respiratory syncytial virus: finding the yellow brick road. Arch Pediatr Adolesc Med.2004;2004 :111– 112
- ↵Hartling L, Wiebe N, Russell K, Patel H, Klassen T. Epinephrine for bronchiolitis. Cochrane Database Syst Rev.2004;1 :CD003123
- ↵Patel H, Platt RW, Lozano JM, Wang EE. Glucocorticoids for acute viral bronchiolitis in hospitalized infants and young children. Cochrane Database Syst Rev.2004;(3):CD004878
- van Woensel JB, Wolfs TF, van Aalderen WM, Brand PL, Kimpen JL. Randomised double blind placebo controlled trial of prednisolone in children admitted to hospital with respiratory syncytial virus bronchiolitis. Thorax.1997;52 :634– 637
- ↵van Woensel JB, van Aalderen WM, de Weerd W, et al. Dexamethasone for treatment of patients mechanically ventilated for lower respiratory tract infection caused by respiratory syncytial virus. Thorax.2003;58 :383– 387
- ↵Cade A, Brownlee KG, Conway SP, et al. Randomised placebo controlled trial of nebulised corticosteroids in acute respiratory syncytial viral bronchiolitis. Arch Dis Child.2000;82 :126– 130
- ↵Perlstein PH, Kotagal UR, Bolling C, et al. Evaluation of an evidence-based guideline for bronchiolitis. Pediatrics.1999;104 :1334– 1341
- ↵Wilson SD, Dahl BB, Wells RD. An evidence-based clinical pathway for bronchiolitis safely reduces antibiotic overuse. Am J Med Qual.2002;17 :195– 199
- ↵Shen Y. Applying the 3M All Patient Refined Diagnosis Related Groups Grouper to measure inpatient severity in the VA. Med Care.2003;41(suppl) :II103–II110
- ↵Wang EE, Law BJ, Boucher FD, et al. Pediatric Investigators Collaborative Network on Infections in Canada (PICNIC) study of admission and management variation in patients hospitalized with respiratory syncytial viral lower respiratory tract infection. J Pediatr.1996;129 :390– 395
- ↵Csonka P, Kaila M, Laippala P, Iso-Mustajarvi M, Vesikari T, Ashorn P. Oral prednisolone in the acute management of children age 6 to 35 months with viral respiratory infection-induced lower airway disease: a randomized, placebo-controlled trial. J Pediatr.2003;143 :725– 730
- Copyright © 2005 by the American Academy of Pediatrics